Thread: How to detect invisible rows caused by the relfrozenxid bug?
Hi, How would one go about detecting whether they've lost rows due to the relfrozenxid? Unfortunately running 'SELECT txid_current() < 2^31' on our DB returns false, and I'm a little bit worried, since we've been seeing some WeirdStuff(tm) lately. We're "only" 200M txids or so past 2^31. Thanks! Regards, Omar
On Thu, Dec 5, 2013 at 4:14 PM, Omar Kilani <omar.kilani@gmail.com> wrote: > Hi, > > How would one go about detecting whether they've lost rows due to the > relfrozenxid? > > Unfortunately running 'SELECT txid_current() < 2^31' on our DB returns > false, and I'm a little bit worried, since we've been seeing some > WeirdStuff(tm) lately. > > We're "only" 200M txids or so past 2^31. > We've been working on coming up with a way to determine this, and I think we're pretty close, so if you can hang tight for a bit, hopefully we can post something. That said, if anyone else has come up with a method, I'd be interested in looking at it. Robert Treat play: xzilla.net work: omniti.com
Hi Robert, Sounds good. Is it safe to upgrade to 9.2.6 (we're on 9.2.5) in the mean time, or should we just leave things untouched? We (unknowingly) got hit by the slave replication bug 2 months ago, too. That was fun. :) Omar On Thu, Dec 5, 2013 at 2:48 PM, Robert Treat <rob@xzilla.net> wrote: > On Thu, Dec 5, 2013 at 4:14 PM, Omar Kilani <omar.kilani@gmail.com> wrote: >> Hi, >> >> How would one go about detecting whether they've lost rows due to the >> relfrozenxid? >> >> Unfortunately running 'SELECT txid_current() < 2^31' on our DB returns >> false, and I'm a little bit worried, since we've been seeing some >> WeirdStuff(tm) lately. >> >> We're "only" 200M txids or so past 2^31. >> > > We've been working on coming up with a way to determine this, and I > think we're pretty close, so if you can hang tight for a bit, > hopefully we can post something. > > That said, if anyone else has come up with a method, I'd be interested > in looking at it. > > > Robert Treat > play: xzilla.net > work: omniti.com